Sakuna Jayasundara

Sakuna Jayasundara

Natural Language Processing and Cyber security researcher, Robotics lover, Chess player

Projects

Object Detection, Tracking and Suspicious Activity Recognition for Maritime Surveillance using Thermal Vision

With this project, we propose a system that is capable not only to detect objects within the surveillance area but also to detect a set of pre-identified suspicious activities happening within the borders. We believe this will be an ideal replacement to the current system available which is to manually detect both objects and classify activities as suspicious or not. With the detection of any such suspicious activities, the system is capable of alerting the relevant authorities in real-time which makes it superior to the available traditional method with an additional benefit of increased safety of security personnel. One key objective of this project is to be able to detect both objects and activities happening at any time of the day. Hence, thermal imagery is used for the development of the models and for real-time detection.
For more info visit my blog.

KORSAL: Key-point Detection based Online Real-Time Spatio-Temporal Action Localization

Real-time and online action localization in a video is a critical yet highly challenging problem. Accurate action localization requires utilization of both temporal and spatial information. Recent attempts achieve this by using computationally intensive 3D CNN architectures or highly redundant two-stream architectures with optical flow, making them both unsuitable for real-time, online applications. To accomplish activity localization under highly challenging real-time constraints, we propose utilizing fast and efficient key-point based bounding box prediction to spatially localize actions. We then introduce a tube-linking algorithm that maintains the continuity of action tubes temporally in the presence of occlusions. Further, we eliminate the need for a two-stream architecture by combining temporal and spatial information into a cascaded input to a single network, allowing the network to learn from both types of information. Temporal information is efficiently extracted using a structural similarity index map as opposed to computationally intensive optical flow. Despite the simplicity of our approach, our lightweight end-to-end architecture achieves state-of-the-art frame-mAP of 74.7% on the challenging UCF101-24 dataset, demonstrating a performance gain of 6.4% over the previous best online methods. We also achieve state-of-the-art video-mAP results compared to both online and offline methods. Moreover, our model achieves a frame rate of 41.8 FPS, which is a 10.7% improvement over contemporary real-time methods.
For more info visit my blog.

FPGA Based Processor for Image Downsampling

The goal of this project is to design a microprocessor specially designed to downsample a given image and the Central Processing Unit (CPU) structure, while simulating it using a Hardware Description Language (HDL) such as Verilog and to implement the given task using a programmable logic device, preferably a Field Programmable Gate Arrays (FPGA).
For more info visit my blog.

BLE Based Indoor Positioning System

Indoor positioning systems enable locating the position of objects or people within buildings. There are different indoor positioning techniques available including Bluetooth Low Energy (BLE), WiFi, UWB, Near Field Communication (NFC), etc. GPS is unreliable for indoor spaces because there is no line of sight with the GPS satellites. This project mainly focuses on an RSSI-based BLE Indoor positioning method. BLE is among the first technologies used for indoor tracking and it is widely used because of its significantly low power consumption and high adoption rate in common electronic appliances. Even though the latencies are high and data transfer capacities are low, it is not much of an issue for our purpose as data is sent periodically and only in small chunks..
For more info visit my blog.